Bug 19815

Summary: [845] Xorg server doesn't start up right
Product: xorg Reporter: Alan Stern <stern>
Component: Driver/intelAssignee: Wang Zhenyu <zhenyu.z.wang>
Status: RESOLVED NOTABUG QA Contact: Xorg Project Team <xorg-team>
Severity: critical    
Priority: medium Keywords: regression
Version: unspecified   
Hardware: x86 (IA32)   
OS: Linux (All)   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
Xorg.conf created by "Xorg -configure"
none
Dmesg log for bootup
none
Xorg.0.log for the single successful startup
none
Xorg.0.log after failed startup (stipple pattern on screen)
none
xorg.conf file for the old server
none
Xorg.0.log from old server after startup and switching to 1280x1024
none
Xorg.0.log from old server after pressing Alt-F7
none
Xorg.0.log from old server after moving mouse none

Description Alan Stern 2009-01-29 13:15:13 UTC
Chipset: 82845G/GL/GE rev 3
Architecture: Intel Pentium 4, 32-bit
Driver: xorg-x11-drv-i810-2.5.0-4.fc10
X server: xorg-x11-server-Xorg-1.5.3-6.fc10
Kernel: 2.6.27.9-159.fc10.i686
Distribution: Fedora 10
Motherboard: Intel D845GEBV2
Connector: VGA

This problem is very difficult to pin down; the symptoms are highly variable.  On only one occasion out of about 20 attempts it worked correctly.

The most common symptom is that the server hangs during startups.  Sometimes it hangs before drawing the stipple pattern on the screen (the screen is totally black); other times it hangs after drawing the stipple pattern but before drawing the mouse cursor.  While hung, the server does not respond to Ctrl-Alt-Fn or Ctrl-Alt-Backspace.  I can't stop the server with a normal "kill" command, although "kill -9" works.

The operating system version doesn't seem to be directly involved.  When I boot with an old 2.6.26 kernel (from Fedora 8) the same thing happens.  And when I boot 2.6.27.9 but start the old X server from Fedora 8 (which uses xorg-x11-drv-i810-2.1.1-7.fc8), it starts up normally.

I will attach xorg.conf, dmesg, and Xorg.0.log files showing what happens.  I will also attach files showing what happens when I try to switch VTs using the old i810 driver -- it may contain a clue.
Comment 1 Alan Stern 2009-01-29 13:17:05 UTC
Created attachment 22361 [details]
Xorg.conf created by "Xorg -configure"
Comment 2 Alan Stern 2009-01-29 13:17:52 UTC
Created attachment 22362 [details]
Dmesg log for bootup
Comment 3 Alan Stern 2009-01-29 13:18:58 UTC
Dmesg log while trying to start the X server:

[drm] Initialized drm 1.1.0 20060810
pci 0000:00:02.0: PCI INT A -> GSI 16 (level, low) -> IRQ 16
pci 0000:00:02.0: setting latency timer to 64
[drm] Initialized i915 1.6.0 20080730 on minor 0
Comment 4 Alan Stern 2009-01-29 13:19:43 UTC
Created attachment 22363 [details]
Xorg.0.log for the single successful startup
Comment 5 Alan Stern 2009-01-29 13:20:59 UTC
Created attachment 22364 [details]
Xorg.0.log after failed startup (stipple pattern on screen)
Comment 6 Alan Stern 2009-01-29 13:22:24 UTC
Additional Xorg.0.log entries added after I pressed Ctrl-Alt-Backspace:

(II) AT Translated Set 2 keyboard: Close
(II) UnloadModule: "evdev"
(II) PS2++ Logitech Mouse: Close
(II) UnloadModule: "evdev"
(II) Macintosh mouse button emulation: Close
(II) UnloadModule: "evdev"
(II) Sleep Button (CM): Close
(II) UnloadModule: "evdev"
(II) Power Button (FF): Close
(II) UnloadModule: "evdev"
Comment 7 Alan Stern 2009-01-29 13:28:21 UTC
The next few attachments show what happens with the old X server when switching VTs.  The sequence is this:

Start the X server and press Ctrl-Alt-KP+ to go into high-resolution mode (I have it set to start in 640x480 -- don't ask why).

Press Ctrl-Alt-F1 to change to a text console VT.

Press Alt-F7 to change back to the X display.  The server is still alive but the display is messed up -- shifted horizontally and vertically in a weird manner.

Move the mouse a little.  At this point the server aborts with an error message in the log.  The screen goes into an unusable display mode, and the only way to do anything is to press Alt-SysRq-r (put the keyboard back in Raw mode) followed by Ctrl-Alt-Delete.  The error message in the log may provide a clue to the problem.
Comment 8 Alan Stern 2009-01-29 13:29:25 UTC
Created attachment 22365 [details]
xorg.conf file for the old server
Comment 9 Alan Stern 2009-01-29 13:30:19 UTC
Created attachment 22366 [details]
Xorg.0.log from old server after startup and switching to 1280x1024
Comment 10 Alan Stern 2009-01-29 13:31:30 UTC
Additional Xorg.0.log entries (old server) after pressing Ctrl-Alt-F1:

(WW) I810(0): Successfully set original devices
(WW) I810(0): Setting the original video mode instead of restoring
        the saved state
(WW) I810(0): Extended BIOS function 0x5f05 failed.
(II) I810(0): BIOS call 0x5f05 not supported, setting refresh with VBE 3 method.
(II) I810(0): xf86UnbindGARTMemory: unbind key 2
(II) I810(0): xf86UnbindGARTMemory: unbind key 1
(II) I810(0): xf86UnbindGARTMemory: unbind key 0
(WW) I810(0): Successfully set original devices (2)
Comment 11 Alan Stern 2009-01-29 13:32:06 UTC
Created attachment 22367 [details]
Xorg.0.log from old server after pressing Alt-F7
Comment 12 Alan Stern 2009-01-29 13:33:17 UTC
Created attachment 22368 [details]
Xorg.0.log from old server after moving mouse

This file contains the errors messages that led the old server to abort.  It's possible that they indicate some sort of hardware problem, but I'm not able to interpret them.
Comment 13 Wang Zhenyu 2009-02-01 18:57:03 UTC
For old chipset like 845G, we might not support well with recent drivers as we had non or very limited testing on those with new drivers. So I think you'd better test with older stable driver, like 2.3.2, 2.4.1, etc.
Comment 14 Alan Stern 2009-02-02 06:55:55 UTC
You don't understand.  I don't want to test an older driver; I want to get the current driver working.

So what should I do?  I'd be happy to test any patches you send.
Comment 15 Michael Fu 2009-02-03 01:19:40 UTC
Can NoAccel option work for you? New driver won't likely bring new feature to you or make things wrong faster..
Comment 16 Alan Stern 2009-02-04 07:28:15 UTC
Adding the NoAccel option solved the problem!  It's working fine now.  Thank you very much.

Closing the bug report.
Comment 17 Alan Stern 2009-02-04 12:56:25 UTC
Adding an extra note...

It turns out that I can leave acceleration enabled (which is good because without acceleration, scrolling is very slow).  The trick is to tell the driver to use the old XAA method instead of the newer EXA.

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.